[LU-3439] User code creating multiple lockfiles with same name

• Status: Closed
• Project: Lustre
• Affects Version/s: Lustre 2.1.5
• Type: Bug
• Priority: Minor
• Reporter: Karl W Schulz (Inactive)
• Assignee: Lai Siyao
• Resolution: Duplicate

Environment

Lustre: 2.1.5
OFED: 1.5.4.1
Kernel: 2.6.32-279.el6.x86_64

• Severity: 3

Description

Not sure if this a known issue already, but we have a user who noticed they can create identical filenames within a single directory via their home-grown locking mechanism. We have worked with them to create a small reproducer which shows the issue on 6 Lustre clients (96 MPI tasks). The net effect of running this test is creation of duplicate filenames:

{{c558-801$ ls -l lockdir/
total 12
---------- 1 karl G-800747 52 Jun 5 08:39 Lockfile.lck
---------- 1 karl G-800747 52 Jun 5 08:39 Lockfile.lck
---------- 1 karl G-800747 52 Jun 5 08:39 Lockfile.lck}}

Tarball attached with user code and example output from our Stampede environment. We can confirm that we do not get the repeat filenames when running the reproducer on Lustre 1.8.6

Thanks.

Issue Links

• is related to: LU-2901 Duplicate filename on the same ldiskf...

Comments

[Kit Westneat (Inactive)]

This looks like LU-2901, we have hit this recently as well but didn't have a reproducer.

[Peter Jones]

Lai

Could you please confirm whether this is a duplicate of LU-2901? If so, are you now able to reproduce this issue and move forward with a fix?

Thanks

Peter

[Lai Siyao]

Hi Kit, could you tell me which file has the duplicate name in the testlog in your tarball? BTW, the reproducer program doesn't close file, which is strange. And is there a way to know when it should stop and quit? I haven't been able to reproduce yet.

[Jodi Levi (Inactive)]

Duplicate of LU-2901